JarCastingpgAdba
An implementation of ADBA, a proposed asynchronous SQL specification, for PostgreSQL.
Asynchronous querying with Future
The primary difference from JDBC is that with ADBA, query execution is managed using futures. When the database returns a result it completes the future, either normally or exceptionally if the query had an error.
Compared to plain JDBC, ADBA, this decouples the network communication to the database from the business logic that produces queries and consumes the results.
This reduces the number of threads needed in the application, since there isn't any need for threads to wait for the database to produce results. This in turn reduces memory footprint and the amount of context switching needed.
Structural improvements over JDBC
-
No parsing of queries in the connection library. This significantly reduces the amount of work the library needs to perform.
-
All queries are sent as prepared statements, this helps mitigate the problem with SQL injection, as it's no longer possible to terminate one query and start a new one in the same statement. It doesn't solve the SQL injection problem totally of course.
-
Query pipelining, the library will send query number 2 over the network without waiting for the result of query number 1 if the arguments to query 2 doesn't depend on query 1. This helps a great deal in reducing query time in environments where there is network latency against the database server.
How does it work
The library's network stack is based on the asynchronous part of the NIO framework.
The programmer interface is a significant rework and upgrade of the old JDBC standard.
The core concepts are DataSources, Sessions and Operations.
DataSourcerepresents the database on the other side of the network, you can startSessions from it.Sessionis a connection to the database, over TCP and optionally TLS in the case of Postgresql.Sessions is where you send in Operations containing SQL queries and get Futures back- Operations represent SQL queries, since there is a large amount of different types of queries, there is also a number of different operations that perform different tasks.
Dependency inclusion
Maven
<dependency>
<groupId>org.postgresql</groupId>
<artifactId>pgadba</artifactId>
<version>0.1.0-ALPHA</version>
</dependency>
Gradle
compile 'org.postgresql:pgadba:0.1.0-ALPHA'
Examples
Creating a DataSource
DataSourceFactory.newFactory("org.postgresql.adba.PgDataSourceFactory")
.builder()
.url("jdbc:postgresql://" + postgres.getContainerIpAddress() + ":" + postgres.getMappedPort(5432)
+ "/" + postgres.getDatabaseName())
.username(postgres.getUsername())
.password(postgres.getPassword())
.build();
This is a straightforward building pattern.
Performing a query
List<Integer> result = new ArrayList<>();
try (Session session = ds.getSession()) {
Submission<List<Integer>> sub = session.<List<Integer>>rowOperation("select $1 as id")
.set("$1", 1, AdbaType.INTEGER)
.collect(() -> result,
(list, row) -> list.add(row.at("id").get(Integer.class)))
.submit();
get10(sub.getCompletionStage());
}
Lets go over this line by line.
- Just a list to hold the result of the operation.
- We start by getting a session in a try-with-resources construct.
getSession()is a convenience function that creates a ConnectionOperation and submits it to the operation stack - Here we create a row operation, queries that returns rows generally want to use row operations, while queries that return the number of rows affected should use countOperations
- Setting of parameters, no parsing of the query is done in the library, so replacing the $1 with the parameter is done by the server in the same way that prepared statements work in regular jdbc.
- The collect call takes a standard java.util.stream.Collector, and have a overloaded default implementation that uses Collector.of().
submit()signals that we are done building theOperationand want to send it off to be executed by the server.- get10 is a helper function that waits for a future to complete, with a timeout of 10s.
Full application example
How can I get involved
This is very much a work in progress. Bail in if interested!
All bug reports, pull requests and suggestions are very welcome.
